On Thu, 09 Oct 2008 15:53:31 -0500, "sauwen" <sauwen.jl@gmail.com>
wrote:

>Hi all.  Does anyone know how to change the width of a feedback comb
>filter's teeth? From the general idea on wikipedia, it seems like it is
>only possible to change the delays and the gain.
>
>Thanks

Hi,
  I'm just thinkin' out loud here, but 
perhaps you could cascade the feedback comb 
with a feedforward comb.

Both combs must have the same delay.  Then 
varying the feedforward value (and sign) of 
alpha may allow you to "adjust" width of the 
cascaded filter's teeth. 

[-Rick-]

Here is the matlab function that I implemented of the filter on the
article


function [b,a]=reson(fn,fs,K,mthd)
%
% [b,a]=reson(fn,fs,K,mthd)
%
% Generates a resonant (single pole) filter.Parameters are:
%
% fn      - desired location of pole (Hz)
% fs      - sampling frequency (Hz)
% K       - Q factor or pole magnitude(depending on witch method used
%           in the mthd parameter described below)
% mthd    - (string) desired method to use. Options are:
%
%             'poleMag' : in this case K specifies the magnitude of
the pole in the filter
%                         (0 < K <1). Notice that the gain might not
be equal to 1 at the resonant
%                         peak. And this method might provide *less*
attenuaton at the sidebands than
%                         method using 'Q2' below.
%
%             'Q1'      : simple method utilizing a second order
feedback system. K specificies
%                         the filter's Q (how narrow the filter is).
This method has significant
%                         drawbacks, use Q2 instead.
%
%             'Q2'     : improved version of Q1 with more symetrical
response and unit gain at desired
%                        pole location (see example).
%
% b       - filter FIR taps
% a       - filter AR taps
%
%
% %Example
% fs=5000;
% fn=50;
% K=50;
% r=0.997;
% [b1,a1]=reson(fn,fs,K,'Q1');
% [b2,a2]=reson(fn,fs,K,'Q2');
% [b3,a3]=reson(fn,fs,r,'poleMag');
%
% [H1,F]=freqz(b1,a1,500,fs);
% hold on;
% [H2,F]=freqz(b2,a2,500,fs);
% [H3,F]= freqz(b3,a3,500,fs);
%  plot(F,20*log10(abs(H1)));hold on;grid on
%  plot(F,20*log10(abs(H2)),'r')
%  plot(F,20*log10(abs(H3)),'g')
% legend('Q1','Q2','PoleMag')

%Written By Ikaro Silva 2008

switch mthd

    case 'poleMag'
        cW = cos(2*pi*fn/fs);
        b=[1];
        a=[1 -2*K*cW K^2];

    case 'Q1'
        % %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
        % Puts a pole exatly on the unit circle
        % at the fn Hz line.

        %Not recommended!!! Use Q2 instead
        cW =2*cos(2*pi*fn/fs);
        Hn=[1 -cW 1];
        b=[1];
        a=[1 K.*Hn];

    case 'Q2'

        %Similar  to 'Q1' but more accurate
        %For details see IEEE SP 2008 (5), pg 113
        beta=1+K;
        f=pi*fn/fs;

        numA=tan(pi/4 - f);
        denA=sin(2*f)+cos(2*f)*numA;
        A=numA/denA;

        b=[1 -2*A A.^2];
        a=[ (beta + K*(A^2)) -2*A*(beta+K) ((A^2)*beta + K)];

end

%%%End of Function%%%%%

Hi,

There was an article that just came out on the IEEE Signal Processing
2008 (5), on the section "DSP Tips& Tricks"
that describes how to design a good variable Q filter (single notch).

hth

On Oct 9, 4:53&#4294967295;pm, "sauwen" <sauwen...@gmail.com> wrote:
> Hi all. &#4294967295;Does anyone know how to change the width of a feedback comb
> filter's teeth? From the general idea on wikipedia, it seems like it is
> only possible to change the delays and the gain.
>
> Thanks